Weigh the evidence before choosing a next step.
Recommendations and causal analysis
Compare possible actions and the evidence supporting them. Distinguish an observed pattern from a measured effect when reviewing what to do next.
Did the collection reminder change payment timing?
Only safe options
Options are ranked only after anything that breaks your rules has been removed. The system won’t accept a setup that ranks first and checks later.
Weighing up goals
Margin, service, cash and risk rarely agree. It shows the trade-off openly and says which goal won, instead of hiding it inside a single score.
Who a nudge will change
It looks for the customers who will act because you contacted them. That list is usually much shorter than the list of customers likely to act anyway.
Checking what worked
It records who saw what and what happened next, and sets out how cause will be tested before claiming anything caused anything.
Measuring the effect of its own suggestions.
A suggestion changes what happens next, which makes it harder to judge whether the suggestion was right. The system keeps track of that instead of ignoring it.
Did the reminder improve collection timing?
Observed change is a starting point for evaluation.
RecommendationsIllustrative records| Measure | Before | After |
|---|---|---|
| Average days late | 14 | 12 |
| Change in customer mix | Not controlled | Needs review |
| Comparison group | Not provided | Needed for a stronger evaluation |
The two-day reduction is an observation, not proof of cause.
Review the comparison design and other changes before attributing the result to the reminder or changing the policy.
- Every suggestion records who saw it, and when.
- Results that followed a plan are marked as influenced by it, so they aren’t counted as proof the plan was right.
- What people actually do adds to the evidence, but never changes your rules.
- Links the system worked out are kept apart from links in your records.
- A pattern can’t be reworded as a cause by the AI that writes answers.
- Claims about cause need a proper test behind them, rather than two figures that move together and a confident sentence.
- Options are ranked only after your rules have removed anything unsafe.
- Your live records are checked again before anything is applied.
A high score doesn’t make something allowed.
This is the most important point on this page. A high rank means an option scored well against your goals. It doesn’t mean it is allowed or safe. Permission comes from your rules and your team’s roles, checked again at the moment anything is applied. Claims about cause are held to a higher bar than forecasts, because acting on a false cause costs more than acting on a weak forecast.
Whenever you ask what to do next.
- “What should we do about this?” — ranked options, each with the goal it favours and what it costs elsewhere.
- “Which of these customers is worth chasing?” — the ones a chase would change, so effort goes where it makes a difference.
- “Did that change work?” — measured against who saw it, rather than a before-and-after that anything could explain.
- “Why this option over that one?” — the trade-off, spelled out.
- “What should I do next here?” — chosen only from options that pass your rules.
Frequently asked questions.
Why check the rules before ranking?
Because ranking an unsafe option puts it in front of someone, and people act on what they see. Checking first means an unsafe option is never shown. The system rejects any setup that does it the other way round.
What is uplift, and why does it matter?
Some customers are likely to pay anyway. Uplift finds the ones who will pay because you chased them. Chasing customers who would have paid anyway is an expensive way to achieve nothing.
Can it claim something caused something else?
Only with a proper test behind it. Two figures that move together and a confident sentence aren’t enough, and the AI that writes answers isn’t allowed to turn a pattern into a cause.
Does a recommendation ever act on its own?
Only for a specific task where you have allowed it. Anything that posts to your books, sends a message or commits money waits for a person with the right permission. Before anything is applied, permissions and records are checked again, and the whole request stops if anything has changed.
Try it on your own records.
Book a demo and ask it a real question about your business. Every answer shows how it was worked out and which records it used.
Book a demo